Characterization of EEG signals using wavelet transform for motor imagination tasks in BCI systems

Brain Computer Interface (BCI) covers an area of special interest, mainly due to the research being conducted to control external devices via thought commands, generating solutions in both motor disability and speech. These applications require the use of signal processing in real time, to be used in devices that help people with this type of disabilities. This paper presents a methodology for feature extraction of electroencephalographic (EEG) signals in motor imagery task of both left and right hand, using the public database BCI Competition 2003. It has been used the wavelet transform for the signal decomposition in the spectral bands of interest (known as brain rhythms). The brain rhythms characterization was conducted through relative energy, variance and standard deviation of the wavelet coefficients. In addition, we conducted the relevance analysis through the fuzzy entropy algorithm, to find the most important features within the training set. We obtained a classification accuracy of up to 98.44% using K-NN and SVM algorithms. The classification results allow inferring that the methodology is appropriate for the recognition of imagination movements in people with motor disabilities and could generate solutions in applications of BCI systems.